Forecasting Value at Risk: An application to Vietnamese market

Abstract

 This paper investigates several approaches to forecast Value at Risk (VaR) in Vietnam. Non-parametric, semi-parametric and parametric methods are used to forecast VaR for returns of Vietnamese stock index. Using several backtesting methods, I find that the GARCH-based model with non-Gaussian conditional return distribution yields the best forecasting performance. This finding warns the popular application of normal distribution to stock returns, which is found to underestimate the potential risk of Vietnamese stock market.